In this edition we’ll be covering…
OpenAI's new ChatGPT Health platform and what it means for 230 million weekly health queries
A complete walkthrough on generating professional video content with Runway 4.5
How to maintain perfect character consistency using Gemini Gems
5 trending AI signals you need to know
3 AI tools to streamline your workflow
And much more…
The Latest in AI
ChatGPT Gets a Medical Badge
OpenAI just launched ChatGPT Health, and it's not just another feature drop—it's a direct play at fixing healthcare's biggest pain points. The company revealed that over 230 million people (wow) already ask health questions on ChatGPT every week…
Here's what's actually new:
ChatGPT Health creates a separate conversation space specifically for health discussions, keeping your medical chats isolated from your standard queries about coding or dinner ideas.
It integrates directly with wellness apps like Apple Health, Function, and MyFitnessPal, pulling in your personal data and medical records to give you contextual health advice.
OpenAI promises not to train their models on Health conversations, addressing the obvious privacy concerns that come with sharing medical information with an AI.
If you start asking health questions in the regular ChatGPT interface, the AI will nudge you to switch to the Health section.
So What?
OpenAI is making a calculated bet that people are desperate enough for accessible healthcare guidance to trust an AI chatbot with their medical concerns. And given their weekly usage numbers, they might be right.
But here's the irony—OpenAI's own terms of service explicitly state that ChatGPT is "not intended for use in the diagnosis or treatment of any health condition." Because at the end of the day, large language models predict the most likely response, not the most correct one. They hallucinate. They don't understand truth.
So while ChatGPT Health might help you prep better questions for your actual doctor or give you a starting point for research, it's essentially a very sophisticated WebMD that talks back. Useful? Maybe. A replacement for real medical advice? OpenAI's lawyers say absolutely not.
Innovation Showcase
Creating Cinematic Content with Runway 4.5
Image from: Runway ML
Runway ML recently dropped Gen 4.5, their latest text-to-video model, and it's making some bold claims about being the best model yet.
How to Use Runway 4.5:
Head to app.runwayml.com/ai-tools/generate.
Each second of Gen 4.5 video costs 25 credits. Their cheapest plan gives you 625 credits, which translates to about 25 seconds of footage. Plan accordingly.
The Prompting Framework:
Runway works best when you follow this structure:
[Camera] shot of [subject/object] [action] in [environment]. [Supporting details]For example:
Side-on static camera shot of a raccoon in zero gravity stealing garbage from a silver trash can. Handheld documentary style, natural camera shake, raw indie film aesthetic.What Actually Works:
Based on real testing, here's what Gen 4.5 handles well:
Physics and motion: The model understands weight, gravity, and realistic movement better than previous versions. An elephant on a see-saw actually lifts the mouse properly.
Cinematic looks: Gen 4.5 excels at dramatic lighting, lens flares, and that "film" quality—think dusk scenes with dust particles catching light.
Fantasy scenarios: If you're creating impossible scenes (like a character painting a ladder that becomes real), Gen 4.5 gets closer than any other model, though it's still not perfect.
What Struggles:
Complex crowd scenes: People in backgrounds still vanish or appear randomly. Consistency across busy frames remains rough.
Precise prompt adherence: Don't expect pixel-perfect execution. The model might give you a running character when you asked for walking.
Pro Tip: Use timestamp prompting for better control:
[00:00 through 00:02] character looks away
[00:02 through 00:03] rapid zoom to eyes
[00:03 through 00:07] slow furrow of browAccording to Runway's official guide, this isn't perfectly precise, but it guides the general sequence effectively.
Get Your Hands Dirty!
How to Lock In Character Consistency with Gemini Gems
Keeping AI-generated characters consistent across multiple images has always been a nightmare. Well, we just cracked that code with Gems.
Here's the exact setup:
Open Gemini and click "Gems" from the left sidebar.
Create a new Gem and give it a clear name and description (something like "Character Consistency Generator").
Upload your reference images to the project knowledge base, these are the images that define your character's look.
Toggle the default tool to "Image" so the Gem knows it's working with visual content.
Paste in this system prompt for instructions:
Purpose and Goals:
- Assist users in generating high-quality images that maintain strict character and style consistency based on specific visual references provided in the Gem knowledge base.
- Ensure that generated images adhere to the artistic parameters, color palettes, and structural features defined by existing assets.
- Provide clear feedback on how consistent a newly generated image is compared to the source material.
Behaviors and Rules:
1) Analysis and Reference:
a) Analyze the visual characteristics of the images stored in the Gem knowledge base, identifying key traits like lighting, line work, and character proportions.
b) Prioritize maintaining these traits across all subsequent image generation tasks.
2) Image Generation Process:
a) When a user requests a new image, reference the 'consistency model' derived from the Gem knowledge base.
b) Generate a prompt that incorporates the specific descriptors needed to reproduce the target character or style accurately.
c) Confirm with the user if the specific pose or setting requested conflicts with the core identity of the character or style reference.
3) Refinement and Feedback:
a) If an image is not perfectly consistent, explain which elements deviated (e.g., 'The eye color was slightly lighter than the reference') and offer to regenerate.
b) Use specific technical terms regarding art styles (e.g., 'isometric', 'cel-shaded', 'cinematic lighting') to ensure clarity.
Overall Tone:
- Professional, technical, and precise.
- Attentive to detail and focused on accuracy.
- Helpful and guiding regarding the limitations of image generation tools.This setup is gold for maintaining brand consistency, creating character-based content series, or building AI influencers. The Gem analyzes your reference images and understands the visual DNA of your character—facial structure, color palette, art style, proportions—then applies those attributes consistently across new generations.
Real-World Applications:
Comic or storyboard characters that appear in multiple scenes
Social media content series with consistent visual identity
Product mockups with the same model or setting
The key is feeding it high-quality reference images that clearly show the characteristics you want to maintain. The more specific your references, the better your results!
Quick Bites
Stay updated with our favorite highlights, dive in for a full flavor of the coverage!
MIT researchers are investigating memorization risks in clinical AI models, demonstrating how AI trained on electronic health records can memorize patient-specific information—potentially violating privacy even with de-identified data.
AMD unveiled new AI PC chips at CES 2026, introducing the Ryzen AI 400 Series with up to 60 TOPs of NPU performance.
In a fascinating display of how easily AI can fool the internet, a Reddit "whistleblower" convinced millions that food delivery companies were exploiting drivers, using AI-generated documents to prove claims. The entire thing was a hoax.
Elon Musk's xAI raised $20 billion in a Series E funding round led by Valor Equity Partners, exceeding their $15 billion target.
Razer announced Project AVA, a 24/7 AI desk companion that appears as a 5.5-inch 3D hologram.
Trending Tools
💎 Gemini Gems - Create custom AI assistants with specialized knowledge bases for consistent character generation, brand management, and personalized workflows.
🎵 StemSplit - AI-powered vocal remover and stem separator that extracts individual audio elements (vocals, drums, bass, melody) from any track for DJs, producers, and content creators.
📹 TopView AI - All-in-one AI video generator that creates professional marketing videos from text, images, or URLs—complete with AI avatars, UGC-style content, and proven viral templates.
Until we Type Again…
Thank you for reading yet another edition of the Digestibly Newsletter!
Reyhan
Early AI leader, advises Fortune 500 companies on AI Development, LLMOps, AI strategy, speaks (for fun) on practical AI. Turns cutting-edge theory into workflows teams can ship today.
Kevin
Ex-AWS SageMaker/Bedrock lead. Shipped infra powering 1M+ devices and $2.5M ARR. Obsessed with high-throughput, low-latency systems— and brings that discipline to every Digestibly release.